A new study conducted by British researchers has unveiled a potentially concerning form of cyber threat where hackers can extract personal data just by listening to someone type. Using a specially-developed deep-learning algorithm, the researchers demonstrated that a hacker could interpret the unique sound of each keystroke to decipher the text being typed, achieving an astonishing 95% accuracy rate.

The findings suggest that obtaining these recordings is surprisingly uncomplicated. A simple cell phone microphone or even the audio function of popular conferencing software like Zoom could be manipulated to capture the sound of typing. Once obtained, the recorded audio can be processed through the algorithm, translating the keystrokes’ acoustic nuances into coherent text.

An AI-enhanced threat: The While the concept of using sound to capture data, termed an “acoustic side-channel attack,” isn’t novel, the integration of artificial intelligence capabilities makes these attacks particularly potent. The researchers caution that the real danger lies in the potential misuse of this technique to intercept sensitive information like passwords and other online credentials.

Attack scenarios: They noted, “The ubiquity of keyboard acoustic emanations not only presents an accessible avenue for cyberattacks but also might lead potential victims to disregard this vulnerability, believing their typing sounds to be inconspicuous.”

The practical application of such an attack presents multiple possibilities. An unsuspecting user could be targeted in a public space, like a coffee shop. Or, with advanced listening equipment, a cybercriminal could eavesdrop even through walls, emphasizing the need for heightened awareness and potential preventative measures against such sophisticated threats.